Why the term about Q(ci) in eqn.(4) in the paper is a constant?

salesforce / PCL

PyTorch code for "Prototypical Contrastive Learning of Unsupervised Representations"

MIT License

570 stars 83 forks source link

Open Tristan-Liu-13 opened 10 months ago

Tristan-Liu-13 commented 10 months ago

I don't understand that you have Q(ci) = p(ci;xi, \theta), then why \sigma Q(ci) log Q(ci) is a constant? Looking forward to your reply.

Volibear1234 commented 6 months ago

I have the same question, i can not understand here.